Improved collaborative filtering algorithm-based service recommendation model

Abstract

The invention relates to an improved collaborative filtering algorithm-based service recommendation model. A method mainly comprises the steps of performing service recommendation based on collaborative filtering of a user; and introducing a Bhattacharyya coefficient and a Jaccard coefficient to improve service recommendation accuracy. By discovering a nearest neighbor of the user, data is mined according to an interest of the user; a target is calculated through a mathematic function; through a scoring matrix, a score enumeration recommendation list is generated according to the matrix; basedon this, the technology is improved based on a conventional collaborative filtering algorithm; and by fusing the Bhattacharyya coefficient and the Jaccard coefficient, accurate service recommendationis performed. A new User-Based Model is formed, and the accuracy of the conventional collaborative filtering algorithm is improved; and the invention belongs to the crossing field of data mining anddeep learning. The personalized recommendation aims to solve the problem of information overload; and by researching interests of different users, most needed resources are actively recommended for the users, so that the conflict between increment of internet information and user demands is better solved.

Description

Technical field [0001] The present invention relates to a service recommendation model based on an improved collaborative filtering algorithm. It mainly discovers the user’s nearest neighbors, mines data according to the user’s interest, calculates the target through a mathematical function, and then generates a score based on a scoring matrix to list a recommendation list , To produce accurate recommendations for target users, which belongs to the intersection of data mining and deep learning. Background technique [0002] With the expansion of the scale of e-commerce, the types and quantities of goods have become larger and larger, and users want to obtain the most suitable service for users among the many online recommendation services, and it takes a lot of time. The fundamental reason is that the service recommendation algorithm does not match the real needs of users. The present invention provides a model that can recommend services more accurately. [0003] Traditional reco...

AI Technical Summary

Problems solved by technology

Ahn et al. proposed the PIP (Proximity-Impact-Popularity) similarity measure, and pointed out that when the cosine similarity has few items jointly rated by users, it is easy to affect the accuracy of the recommendation, and the similarity calculation is larger than the actual deviation.

Later, more defects were continuously exposed. When there is little user data, User-Based recommendations are difficult to produce convincing, and the timeliness will not be easily changed. When there is a lot of user data, the cost of User-Based calculation matrix will be very high. Big

The Item-Based recommendation will change immediately as the user's behavior changes. Although it is real-time, the amount of calculation is quite large when the data is small, and there is no way to update the item similarity table offline. Products are recommended to users

At the same time, the main disadvantage is that user ratings have a great impact on the system, regardless of the difference in ratings between users, they all have a high similarity

And if the algorithm does not consider the absolute value of the score, it will directly make it difficult for users to distinguish

When the User-Item matrix scoring data is sparse, it is difficult to judge the user's interest, making the accuracy low

Images